1. Field of the Invention
The present invention relates to an image processing apparatus and image processing method which performs image coding processing to code an image.
2. Description of Related Art
A method using motion-compensated prediction has been known as a method of coding and compressing a moving image. The motion-compensated prediction is to detect a motion vector of a moving image by image block matching which detects an image area in a reference image similar to an image area in a part of an image.
On the other hand, fractal image coding which uses a structural feature of an image of self-similarity to compress the image has been known as a coding technique at high compression rate. This fractal image coding is to search a domain block in a specified area similar to a range block which is a part of an image by rotation or the like of the range block. This processing also uses the image block matching. Specifically, an image is divided into non-overlapping range blocks (unit blocks of R×R pixels) firstly, and a domain block (a block of D×D pixels, D>R) larger than the range block is also cut out of the image. This range block is subjected to an affine transformation (scaling, rotation, mirroring and the like) so as to search a domain block having the most similar pattern to the range block. Thereafter, information of the relation between these blocks is stored as a code.
Here, searching a domain block similar to a range block takes a long time. In order to reduce the time, an image coding method has been developed, which parameterizes range blocks and domain blocks previously and extract domain blocks having the same parameter as a certain range block as candidates for coding (for example, see JP 09-275565A).
However, in the method of JP 09-275565A or the like, a parameter for the narrowing down is not invariant to an affine transformation. Therefore, it has been problematic that such method does not work sufficiently for improving the processing which uses an affine transformation.